1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
|
Subject: [PATCH] cfg80211: fix deadlock
From: Johannes Berg <johannes@sipsolutions.net>
To: John Linville <linville@tuxdriver.com>
Cc: linux-wireless <linux-wireless@vger.kernel.org>,
Christian Lamparter <chunkeey@web.de>
Content-Type: text/plain
Date: Sun, 16 Aug 2009 13:32:38 +0200
Message-Id: <1250422358.17522.0.camel@johannes.local>
Mime-Version: 1.0
X-Mailer: Evolution 2.27.90
Content-Transfer-Encoding: 7bit
Sender: linux-wireless-owner@vger.kernel.org
Precedence: bulk
List-ID: <linux-wireless.vger.kernel.org>
X-Mailing-List: linux-wireless@vger.kernel.org
When removing an interface with nl80211, cfg80211 will
deadlock in the netdev notifier because we're already
holding rdev->mtx and try to acquire it again to verify
the scan has been done.
This bug was introduced by my patch
"cfg80211: check for and abort dangling scan requests".
To fix this, move the dangling scan request check into
wiphy_unregister(). This will not be able to catch all
cases right away, but if the scan problem happens with
a manual ifdown or so it will be possible to remedy it
by removing the module/device.
Additionally, add comments about the deadlock scenario.
Reported-by: Christian Lamparter <chunkeey@web.de>
Signed-off-by: Johannes Berg <johannes@sipsolutions.net>
---
net/wireless/core.c | 32 +++++++++++++++++++-------------
1 file changed, 19 insertions(+), 13 deletions(-)
--- a/net/wireless/core.c
+++ b/net/wireless/core.c
@@ -586,9 +586,15 @@ void wiphy_unregister(struct wiphy *wiph
* get to lock contention here if userspace issues a command
* that identified the hardware by wiphy index.
*/
- mutex_lock(&rdev->mtx);
- /* unlock again before freeing */
- mutex_unlock(&rdev->mtx);
+ cfg80211_lock_rdev(rdev);
+
+ if (WARN_ON(rdev->scan_req)) {
+ rdev->scan_req->aborted = true;
+ ___cfg80211_scan_done(rdev);
+ }
+
+ cfg80211_unlock_rdev(rdev);
+ flush_work(&rdev->scan_done_wk);
cfg80211_debugfs_rdev_del(rdev);
@@ -603,9 +609,7 @@ void wiphy_unregister(struct wiphy *wiph
mutex_unlock(&cfg80211_mutex);
- flush_work(&rdev->scan_done_wk);
cancel_work_sync(&rdev->conn_work);
- kfree(rdev->scan_req);
flush_work(&rdev->event_work);
}
EXPORT_SYMBOL(wiphy_unregister);
@@ -653,6 +657,11 @@ static int cfg80211_netdev_notifier_call
switch (state) {
case NETDEV_REGISTER:
+ /*
+ * NB: cannot take rdev->mtx here because this may be
+ * called within code protected by it when interfaces
+ * are added with nl80211.
+ */
mutex_init(&wdev->mtx);
INIT_LIST_HEAD(&wdev->event_list);
spin_lock_init(&wdev->event_lock);
@@ -730,13 +739,11 @@ static int cfg80211_netdev_notifier_call
#endif
break;
case NETDEV_UNREGISTER:
- cfg80211_lock_rdev(rdev);
-
- if (WARN_ON(rdev->scan_req && rdev->scan_req->dev == dev)) {
- rdev->scan_req->aborted = true;
- ___cfg80211_scan_done(rdev);
- }
-
+ /*
+ * NB: cannot take rdev->mtx here because this may be
+ * called within code protected by it when interfaces
+ * are removed with nl80211.
+ */
mutex_lock(&rdev->devlist_mtx);
/*
* It is possible to get NETDEV_UNREGISTER
@@ -755,7 +762,6 @@ static int cfg80211_netdev_notifier_call
#endif
}
mutex_unlock(&rdev->devlist_mtx);
- cfg80211_unlock_rdev(rdev);
break;
case NETDEV_PRE_UP:
if (!(wdev->wiphy->interface_modes & BIT(wdev->iftype)))
|
